Disclosure of Invention
The embodiment of the disclosure provides an image acquisition method, an image acquisition device, electronic equipment and a storage medium, which are used for reducing redundancy in displaying dishes and improving the use experience of a user.
According to a first aspect of embodiments of the present disclosure, there is provided an image acquisition method including:
acquiring a plurality of dish images of a merchant and a merchant main map;
acquiring a dish title text corresponding to each dish image and a main picture title text corresponding to the merchant main picture;
acquiring the dish features in each dish image and the main picture features in the main picture of the merchant;
determining a dish representation vector of each dish image based on the dish title text and the dish features of each dish image;
determining a host graph characterization vector of the merchant host graph based on the host graph title text and the host graph feature of the merchant host graph;
clustering the plurality of dish images based on the dish characterization vectors to generate a plurality of image clusters;
acquiring an estimated click rate corresponding to each dish image based on each dish characterization vector, the main graph characterization vector and a preset click rate estimation model;
and acquiring a dish image with the highest estimated click rate from each image cluster as a dish image to be displayed.
Optionally, the determining a dish characterization vector for each dish image based on the dish title text and the dish features of each dish image includes:
acquiring a first characterization vector corresponding to the dish features in each dish image;
acquiring a first text representation vector corresponding to each dish image based on each dish title text;
and combining the first representation vector of each dish image with the first text representation vector to generate a dish representation vector of each dish image.
Optionally, the determining a principal characterization vector for the merchant principal based on the principal title text and the principal feature of the merchant principal comprises:
acquiring a second characterization vector corresponding to the main feature;
acquiring a second text representation vector corresponding to the merchant main graph based on the main graph title text;
combining the second token vector and the second text token vector to generate the master token vector.
Optionally, the clustering the plurality of dish images based on each of the dish characterizing vectors to generate a plurality of image class clusters includes:
clustering a plurality of dish characterization vectors by adopting a preset clustering algorithm to generate a plurality of image clusters; each image cluster comprises at least one dish image.
Optionally, the obtaining of the estimated click rate corresponding to each dish image based on each dish characterization vector, the main graph characterization vector, and a preset click rate estimation model includes:
acquiring similarity corresponding to each dish image based on each dish characterization vector and the main picture characterization vector;
and inputting each similarity into the preset click rate evaluation model, and outputting the estimated click rate corresponding to each dish image.
Optionally, the obtaining a similarity corresponding to each dish image based on each dish characterization vector and the main picture characterization vector includes:
and calculating cosine similarity between each dish characterization vector and the main picture characterization vector, and taking each cosine similarity as the similarity of each dish image.
Optionally, after the obtaining a dish image with the highest estimated click rate from each image class cluster as the dish image to be displayed, the method further includes:
and fusing the merchant main graph and each dish image to be displayed to generate a display image of the merchant on a specified service platform.
According to a second aspect of embodiments of the present disclosure, there is provided an image acquisition apparatus including:
the merchant main map acquisition module is used for acquiring a plurality of dish images of a merchant and a merchant main map;
a title text acquisition module, configured to acquire a title text of a dish corresponding to each dish image and a main picture title text corresponding to the main picture of the merchant;
the vegetable main map feature acquisition module is used for acquiring vegetable features in each vegetable image and main map features in the merchant main map;
the dish representation vector determination module is used for determining a dish representation vector of each dish image based on the dish title text and the dish features of each dish image;
a principal graph characterization vector determination module for determining a principal graph characterization vector for the merchant principal graph based on the principal graph title text and the principal graph features of the merchant principal graph;
the image cluster generating module is used for clustering the plurality of dish images based on the dish characterization vectors to generate a plurality of image clusters;
the estimated click rate obtaining module is used for obtaining estimated click rates corresponding to the dish images based on the dish characterization vectors, the main picture characterization vector and a preset click rate estimation model;
and the to-be-displayed image acquisition module is used for acquiring a dish image with the highest estimated click rate from each image cluster as the to-be-displayed dish image.
Optionally, the dish characterization vector determination module includes:
the first characterization vector acquisition submodule is used for acquiring a first characterization vector corresponding to the dish features in each dish image;
the first text vector acquisition submodule is used for acquiring a first text representation vector corresponding to each dish image based on each dish title text;
and the dish representation vector generation submodule is used for combining the first representation vector and the first text representation vector of each dish image to generate the dish representation vector of each dish image.
Optionally, the principal characterization vector determination module includes:
the second characterization vector acquisition submodule is used for acquiring a second characterization vector corresponding to the main feature;
the second text vector acquisition sub-module is used for acquiring a second text representation vector corresponding to the merchant main map based on the main map title text;
and the main picture characterization vector generation sub-module is used for combining the second characterization vector and the second text characterization vector to generate the main picture characterization vector.
Optionally, the image class cluster generating module includes:
the image cluster generation submodule is used for clustering the dish characterization vectors by adopting a preset clustering algorithm to generate a plurality of image clusters; each image cluster comprises at least one dish image.
Optionally, the estimated click rate obtaining module includes:
the similarity obtaining submodule is used for obtaining the similarity corresponding to each dish image based on each dish characterization vector and the main picture characterization vector;
and the estimated click rate output submodule is used for inputting each similarity to the preset click rate evaluation model and outputting the estimated click rate corresponding to each dish image.
Optionally, the similarity obtaining sub-module includes:
and the cosine similarity operator module is used for calculating cosine similarity between each dish characterization vector and the main picture characterization vector and taking each cosine similarity as the similarity of each dish image.
Optionally, the method further comprises:
and the display image generation module is used for fusing the merchant main image and each vegetable image to be displayed to generate a display image of the merchant on a specified service platform.
According to a third aspect of embodiments of the present disclosure, there is provided an electronic apparatus including:
a processor, a memory and a computer program stored on the memory and executable on the processor, the processor implementing the image acquisition method of any of the above when executing the program.
According to a fourth aspect of embodiments of the present disclosure, there is provided a computer-readable storage medium having instructions which, when executed by a processor of an electronic device, enable the electronic device to perform any one of the image acquisition methods described above.
The embodiment of the disclosure provides an image obtaining scheme, which includes obtaining a plurality of vegetable images of a merchant and a merchant main map, obtaining a vegetable title text corresponding to each vegetable image and a main map title text corresponding to the merchant main map, obtaining vegetable features in each vegetable image and main map features in the merchant main map, determining a vegetable characterization vector of each vegetable image based on the vegetable title text and the vegetable features of each vegetable image, determining a main map characterization vector of the merchant main map based on the main map title text and the main map features of the merchant main map, performing clustering processing on the plurality of vegetable images based on the vegetable characterization vectors, generating a plurality of image cluster, obtaining an estimated click rate corresponding to each vegetable image based on each vegetable characterization vector, each main map characterization vector and a preset click rate estimation model, obtaining a vegetable image with the highest estimated click rate from each image cluster, as an image of the dish to be displayed. According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image type clusters, and the dish images with the highest estimated click rate are selected from each image type cluster to serve as the displayed dishes of the merchants, so that the redundancy of the displayed dishes can be avoided, and the use experience of users can be improved.
Detailed Description
Technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are some, but not all, of the embodiments of the present disclosure. All other embodiments, which can be obtained by a person skilled in the art without making creative efforts based on the embodiments of the present disclosure, belong to the protection scope of the embodiments of the present disclosure.
Example one
Referring to fig. 2, a flowchart illustrating steps of an image acquisition method provided by an embodiment of the present disclosure is shown, where the image acquisition method may include the following steps:
step 101: and acquiring a plurality of dish images of the commercial tenant and a commercial tenant main map.
The embodiment of the disclosure can be applied to a scene of displaying the recommended dish image of the merchant for a specified service platform (such as a American group takeout platform).
In this disclosure, a merchant refers to a catering merchant.
The dish image is an image of a dish operated by a merchant, and the merchant main graph is an image containing characters such as a merchant name, a merchant Logo, merchant preferential activities and the like.
In some examples, the dish image and merchant host may be uploaded by the merchant onto a specified business platform. Specifically, a merchant can shoot a gate by using equipment such as a camera to obtain a gate map, process the gate map of the merchant, add characters such as a merchant name, a merchant Logo and merchant preferential activities to generate a merchant main map, take pictures of characteristic dishes of the merchant main map to obtain a plurality of dish images, and upload the obtained merchant main map and dish images to a specified service platform.
In some examples, the menu image and the merchant main map may be images of the merchant collected by business personnel arranged on a designated business platform, for example, the business personnel take a live shot at the location of the merchant and perform post-processing to obtain a plurality of menu images and merchant main maps of the merchant.
It is to be understood that the foregoing examples are merely examples listed for better understanding of technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitation to the embodiments of the present disclosure, and in a specific implementation, the dish image of the merchant and the merchant main graph may also be obtained from a comment area, such as an uploaded dish image and merchant main graph when the merchant promotes other dishes in the comment area, and the like.
After acquiring the plurality of vegetable images of the merchant and the merchant main map, step 102 and step 103 are executed.
Step 102: and acquiring a dish title text corresponding to each dish image and a main picture title text corresponding to the merchant main picture.
The dish title text refers to the title text corresponding to each dish image, such as the title text of spicy hot pots, spicy hot shrimps and the like.
The main map title text refers to the title text corresponding to the main map of the merchant, such as the title text of "xxx crayfish", "xxx sheep soup house" and the like.
In the present disclosure, the dish title text of each dish image and the subject title text of the merchant main map may be texts input by the merchant for each dish image and one merchant main map. The text may also be a text input by a service person of a specific service platform for each menu image and one merchant main map, and specifically, the text may be determined according to service requirements, which is not limited in this embodiment of the disclosure.
Step 103: and acquiring the dish features in each dish image and the main picture features in the main picture of the merchant.
The dish feature refers to a feature of the dish included in the dish image, and for example, as shown in fig. 1, the dish feature included in the display image is "shrimp".
The main feature is a feature included in the merchant main graph, and as shown in fig. 1, Logo and the like of the merchant in the display image are the main feature.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
After obtaining a plurality of vegetable images of the merchant and one merchant main map, image recognition processing may be performed on each vegetable image and the merchant main map by using an image recognition technology to obtain vegetable features in each vegetable image and main map features in the merchant main map.
Of course, in a specific implementation, the dish feature of each dish image and the main feature in the main graph of the merchant may also be obtained in other manners, such as manual entry, and the like, and specifically, the method may be determined according to business requirements, and the embodiment of the present disclosure does not limit this.
After the dish title text corresponding to each dish image and the dish features in each dish image, and the main picture title text corresponding to the merchant main picture and the main picture features in the merchant main picture are obtained, step 104 is executed.
Step 104: and determining a dish characterization vector of each dish image based on the dish title text and the dish features of each dish image.
Step 105: determining a host graph characterization vector for the merchant host graph based on the host graph title text and the host graph features of the merchant host graph.
The dish characterization vector is obtained by combining a dish title text and dish features of a dish image.
The main graph characterization vector is used for converting the main graph characteristics and the main graph title text in the main graph of the merchant into vectors for representation.
After obtaining a plurality of vegetable images of the merchant and one merchant main map, an inclusion V4 model can be adopted to extract the vegetable images and the image characteristics (namely the vegetable characteristics and the main map characteristics) of the merchant main map, and the extracted image characteristics are compressed through a network layer Avg Pooling, so that the vector representation of each vegetable image can be obtained, and the vector representation of the merchant main map can be obtained.
And then, obtaining a text representation vector corresponding to each dish image according to the dish title text corresponding to each dish image, and obtaining a text representation vector corresponding to the merchant main picture according to the main picture title text corresponding to the merchant main picture.
And combining the characterization vector of the dish features of each dish image and the characterization vector of the dish title text to obtain the dish characterization vector of each dish image.
And combining the characterization vector of the main character of the merchant main graph and the characterization vector of the main character title text to obtain the main graph characterization vector of the merchant main graph.
The specific manner of obtaining the dish characterization vector corresponding to each dish image and the main host characterization vector corresponding to the main host of the merchant will be described in detail in the following embodiment two, which is not repeated herein in the embodiments of the present disclosure.
After the dish characterization vector corresponding to each dish image and the main graph characterization vector corresponding to the merchant main graph are obtained, step 106 is executed.
Step 106: and clustering the plurality of dish images based on the dish characterization vectors to generate a plurality of image clusters.
The image cluster refers to a cluster obtained by classifying the dish images, wherein the dish features of the dish images in each image cluster are similar, so that the dish features of the two dish images in different image clusters are different.
After the dish characterizing vector of each dish image in the plurality of dish images is obtained, the plurality of dish images can be clustered according to the dish characterizing vector of each dish image, so that a plurality of image clusters can be generated, and specifically, a preset clustering algorithm can be adopted to perform clustering on the characterizing vectors of the plurality of dish images, so that a plurality of image clusters can be generated.
The detailed process of the clustering process will be described in the following embodiments, which are not described herein again in this disclosure.
After clustering the plurality of dish images based on the dish characterizing vectors to generate a plurality of image class clusters, step 107 is performed.
Step 107: and acquiring the estimated click rate corresponding to each dish image based on each dish characterization vector, the main graph characterization vector and a preset click rate estimation model.
The estimated click rate refers to the click rate of each dish image obtained through evaluation.
The preset click rate estimation model is a model used for evaluating the click rate of each dish image.
After the vegetable characterization vector of each vegetable image and the main map characterization vector of the main map of the merchant are obtained, the vegetable characterization vector of each vegetable image and the main map characterization vector of the main map of the merchant can be input into a preset click rate estimation model, so that the estimated click rate corresponding to each vegetable image can be obtained.
After obtaining the estimated click rate corresponding to each dish image, step 108 is executed.
Step 108: and acquiring a dish image with the highest estimated click rate from each image cluster as a dish image to be displayed.
After the estimated click rates corresponding to the plurality of dish images are obtained, a dish image with the highest estimated click rate can be obtained from each image cluster according to the obtained image clusters to serve as a dish image to be displayed, for example, the image clusters comprise a cluster A, a cluster B and a cluster C, the dish image with the highest estimated click rate in the cluster A is an image a, the dish image with the highest estimated click rate in the cluster B is an image B, the dish image with the highest estimated click rate in the cluster C is an image C, and the image a, the image B and the image C are used as the dish images to be displayed.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image clusters, and the dish image with the highest estimated click rate is selected from each image cluster to serve as the displayed dish of the merchant, so that redundancy of the displayed dish can be avoided.
The image obtaining method provided by the embodiment of the disclosure obtains a plurality of vegetable images and a merchant main map of a merchant, obtains a vegetable title text corresponding to each vegetable image and a main map title text corresponding to the merchant main map, obtains vegetable features in each vegetable image and main map features in the merchant main map, determines a vegetable characterization vector of each vegetable image based on the vegetable title text and the vegetable features of each vegetable image, determines a main map characterization vector of the merchant main map based on the main map title text and the main map features of the merchant main map, performs clustering processing on the plurality of vegetable images based on the vegetable characterization vectors to generate a plurality of image clusters, obtains an estimated click rate corresponding to each vegetable image based on each vegetable characterization vector, the main map characterization vector and a preset click rate estimation model, obtains a vegetable image with the highest estimated click rate from each image cluster, as an image of the dish to be displayed. According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image type clusters, and the dish images with the highest estimated click rate are selected from each image type cluster to serve as the displayed dishes of the merchants, so that the redundancy of the displayed dishes can be avoided, and the use experience of users can be improved.
Example two
Referring to fig. 3, a flowchart illustrating steps of an image acquisition method provided by an embodiment of the present disclosure is shown, where the image acquisition method may include the following steps:
step 201: and acquiring a plurality of dish images of the commercial tenant and a commercial tenant main map.
The embodiment of the disclosure can be applied to a scene of displaying the recommended dish image of the merchant for a specified service platform (such as a American group takeout platform).
In this disclosure, a merchant refers to a catering merchant.
The dish image is an image of a dish operated by a merchant, and the merchant main graph is an image containing characters such as a merchant name, a merchant Logo, merchant preferential activities and the like.
In some examples, the dish image and merchant host may be uploaded by the merchant onto a specified business platform. Specifically, a merchant can shoot a gate by using equipment such as a camera to obtain a gate map, process the gate map of the merchant, add characters such as a merchant name, a merchant Logo and merchant preferential activities to generate a merchant main map, take pictures of characteristic dishes of the merchant main map to obtain a plurality of dish images, and upload the obtained merchant main map and dish images to a specified service platform.
In some examples, the menu image and the merchant main map may be images of the merchant collected by business personnel arranged on a designated business platform, for example, the business personnel take a live shot at the location of the merchant and perform post-processing to obtain a plurality of menu images and merchant main maps of the merchant.
It is to be understood that the foregoing examples are merely examples listed for better understanding of technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitation to the embodiments of the present disclosure, and in a specific implementation, the dish image of the merchant and the merchant main graph may also be obtained from a comment area, such as an uploaded dish image and merchant main graph when the merchant promotes other dishes in the comment area, and the like.
After obtaining the plurality of images of the dish and the merchant main map of the merchant, step 202 is performed.
Step 202: and acquiring a dish title text corresponding to each dish image and a main picture title text corresponding to the merchant main picture.
The dish title text refers to the title text corresponding to each dish image, such as the title text of spicy hot pots, spicy hot shrimps and the like.
The main map title text refers to the title text corresponding to the main map of the merchant, such as the title text of "xxx crayfish", "xxx sheep soup house" and the like.
In the present disclosure, the dish title text of each dish image and the subject title text of the merchant main map may be texts input by the merchant for each dish image and one merchant main map. The text may also be a text input by a service person of a specific service platform for each menu image and one merchant main map, and specifically, the text may be determined according to service requirements, which is not limited in this embodiment of the disclosure.
Step 203: and acquiring the dish features in each dish image and the main picture features in the main picture of the merchant.
The dish feature refers to a feature of the dish included in the dish image, and for example, as shown in fig. 1, the dish feature included in the display image is "shrimp".
The main feature is a feature included in the merchant main graph, and as shown in fig. 1, Logo and the like of the merchant in the display image are the main feature.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
After obtaining a plurality of vegetable images of the merchant and one merchant main map, image recognition processing may be performed on each vegetable image and the merchant main map by using an image recognition technology to obtain vegetable features in each vegetable image and main map features in the merchant main map.
Of course, in a specific implementation, the dish feature of each dish image and the main feature in the main graph of the merchant may also be obtained in other manners, such as manual entry, and the like, and specifically, the method may be determined according to business requirements, and the embodiment of the present disclosure does not limit this.
After the dish title text corresponding to each dish image and the dish features in each dish image, and the main picture title text corresponding to the merchant main picture and the main picture features in the merchant main picture are obtained, step 204 is executed.
Step 204: and acquiring a first characterization vector corresponding to the dish features in each dish image.
The first characterization vector is to convert the dish features of the dish images into vectors for representation, for example, the dish features of each dish image are compressed by the Avg Pooling network layer to finally obtain a 154-dimensional vector representation, and the 154-dimensional vector representation is the first characterization vector.
After the dish features in each dish image are obtained, the dish features can be subjected to vector representation, so that a first characterization vector corresponding to the dish features in each dish image can be obtained.
After the first characterization vector corresponding to the dish features in each dish image is obtained, step 205 is performed.
Step 205: and acquiring a first text representation vector corresponding to each dish image based on each dish title text.
The first text representation vector is used for converting the dish title text of the dish image into a vector for representation.
The way to obtain the first text characterization vector may be: firstly, word segmentation is carried out on a dish title text, then a PVDM algorithm realized based on a FastText tool is adopted to extract vector representation of dish text information, the vector representation of each dish title text can be directly obtained through the algorithm, for example, the dish title text of each dish image is converted into 154-dimensional vector representation, and the 154-dimensional vector representation is a first text representation vector.
After the first text representation vector corresponding to each dish image is obtained based on each dish title text, step 206 is executed.
Step 206: and combining the first representation vector of each dish image with the first text representation vector to generate a dish representation vector of each dish image.
The dish representation vector is a representation vector obtained by combining a first representation vector of a dish image and a first text representation vector.
After obtaining the first token vector and the first text token vector of each dish image, the first token vector and the first text token vector of each dish image may be combined, so that the dish token vector of each dish image may be generated, for example, in the above example, the obtained first token vector is a 154-dimensional vector representation, the first text token vector is also a 154-dimensional vector representation, and a 308-dimensional vector representation may be obtained by combining the first token vector and the first text token vector, where the 308-dimensional vector representation is the dish token vector.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
Step 207: and acquiring a second characterization vector corresponding to the main feature.
The second characterization vector is to convert the principal graph feature of the merchant principal graph into a vector for representation, for example, the principal graph feature of the merchant principal graph is compressed by the network layer Avg Pooling to finally obtain a 154-dimensional vector representation, and the 154-dimensional vector representation is the second characterization vector.
After obtaining the principal graph feature in the merchant principal graph, the principal graph feature in the merchant principal graph can be converted into a vector representation, that is, a second characterization vector corresponding to the principal graph feature in the merchant principal graph is obtained.
After the second token vector corresponding to the main feature is obtained, step 208 is performed.
Step 208: and acquiring a second text characterization vector corresponding to the merchant main graph based on the main graph title text.
The second text characterization vector is used for converting the main host title text of the main host of the merchant into a vector for representation.
The way to obtain the second text characterization vector may be: firstly, segmenting words of a subject title text, and then extracting vector representation of the main picture title text by adopting a PVDM algorithm realized based on a FastText tool, wherein the algorithm can directly obtain the vector representation of the main picture title text, for example, the main picture title text of the main picture image is converted into 154-dimensional vector representation, and the 154-dimensional vector representation is a second text representation vector.
After obtaining a second text characterization vector corresponding to the merchant host based on the host title text, step 209 is performed.
Step 209: combining the second token vector and the second text token vector to generate the master token vector.
The main graph characterization vector is a characterization vector obtained by combining a second characterization vector of the main graph of the merchant with a second text characterization vector.
For example, in the above example, the obtained second token vector is a 154-dimensional vector representation, the obtained second text token vector is also a 154-dimensional vector representation, and the second token vector and the second text token vector are combined to obtain a 308-dimensional vector representation, where the 308-dimensional vector representation is the main graph token vector.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
Step 210: clustering a plurality of dish characterization vectors by adopting a preset clustering algorithm to generate a plurality of image clusters; each image cluster comprises at least one dish image.
The preset clustering algorithm is an algorithm for classifying a plurality of dish images, and may be a Keans algorithm or the like, and specifically, may be determined according to business requirements, which is not limited in the embodiments of the present disclosure.
The image cluster refers to a cluster obtained by classifying the dish images, wherein the dish features of the dish images in each image cluster are similar, so that the dish features of the two dish images in different image clusters are different.
After the dish characterizing vectors corresponding to the plurality of dish images are obtained, the plurality of dish characterizing vectors can be input into a preset clustering algorithm, and the preset clustering algorithm is used for clustering the plurality of dish images, so that a plurality of image clusters can be generated. At least one dish image is contained in each image cluster.
After the preset clustering algorithm is adopted to perform clustering processing on the plurality of dish characterization vectors to generate a plurality of image class clusters, step 211 is executed.
Step 211: and acquiring the corresponding similarity of each dish image based on each dish characterization vector and the main picture characterization vector.
After the vegetable characterization vector of each vegetable image and the main map characterization vector of the main map of the merchant are obtained, the similarity can be calculated according to each vegetable characterization vector and the main map characterization vector, namely the cosine similarity between each vegetable characterization vector and the main map characterization vector is calculated, and each cosine similarity is used as the similarity between each vegetable image and the main map of the merchant.
After obtaining the corresponding similarity of each dish image based on each dish characterization vector and the main graph characterization vector, step 212 is performed.
Step 212: and inputting each similarity into the preset click rate evaluation model, and outputting the estimated click rate corresponding to each dish image.
The estimated click rate refers to the click rate of each dish image obtained through evaluation.
The preset click rate estimation model is a model used for evaluating the click rate of each dish image.
After the similarity corresponding to each dish image is obtained, the similarity value can be input into a preset click rate evaluation model, and the estimated click rate of each dish image is output by the preset click rate evaluation model.
Certainly, in a specific implementation, discrete features (such as merchant identifiers, dish identifiers, interested dish identifiers and the like) of each dish image, statistical features (that is, statistical features of the user and the dishes, such as the praise times and the like of the dishes), dish feature characterization vectors of the dish images, text characterization vectors and the like may be input to the preset click rate evaluation model together, so as to calculate the estimated click rate of each dish image according to the multidimensional features.
After inputting each similarity to the preset click rate evaluation model and outputting the estimated click rate corresponding to each dish image, step 213 is executed.
Step 213: and acquiring a dish image with the highest estimated click rate from each image cluster as a dish image to be displayed.
After the estimated click rates corresponding to the plurality of dish images are obtained, a dish image with the highest estimated click rate can be obtained from each image cluster according to the obtained image clusters to serve as a dish image to be displayed, for example, the image clusters comprise a cluster A, a cluster B and a cluster C, the dish image with the highest estimated click rate in the cluster A is an image a, the dish image with the highest estimated click rate in the cluster B is an image B, the dish image with the highest estimated click rate in the cluster C is an image C, and the image a, the image B and the image C are used as the dish images to be displayed.
It is to be understood that the above examples are only examples set forth for a better understanding of the technical solutions of the embodiments of the present disclosure, and are not to be taken as the only limitations on the embodiments of the present disclosure.
According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image clusters, and the dish image with the highest estimated click rate is selected from each image cluster to serve as the displayed dish of the merchant, so that redundancy of the displayed dish can be avoided.
After a dish image with the highest estimated click rate is obtained from each image class cluster and used as the dish image to be displayed, step 214 is executed.
Step 214: and fusing the merchant main graph and each dish image to be displayed to generate a display image of the merchant on a specified service platform.
After the plurality of vegetable images to be displayed are obtained, the plurality of vegetable images to be displayed and the main business computer can be fused, so that the display image of the business on the appointed business platform can be generated, for example, as shown in fig. 1, the display image of the business on the appointed business platform can include three displayed vegetables and Logo, preferential information and the like of the business.
The image obtaining method provided by the embodiment of the disclosure obtains a plurality of vegetable images and a main merchant main map of a merchant, obtains a vegetable title text corresponding to each vegetable image and a main map title text corresponding to the merchant main map, obtains vegetable features in each vegetable image and main map features in the merchant main map, determines a vegetable characterization vector of each vegetable image based on the vegetable title text and the vegetable features of each vegetable image, determines a main map characterization vector of the merchant main map based on the main map title text and the main map features of the merchant, performs clustering processing on the plurality of vegetable images based on the vegetable characterization vectors to generate a plurality of image clusters, obtains a click rate corresponding to each vegetable image based on each vegetable characterization vector, the main map characterization vector and a preset click rate prediction model, obtains a vegetable image with the highest predicted click rate from each image cluster, as an image of the dish to be displayed. According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image type clusters, and the dish images with the highest estimated click rate are selected from each image type cluster to serve as the displayed dishes of the merchants, so that the redundancy of the displayed dishes can be avoided, and the use experience of users can be improved.
EXAMPLE III
Referring to fig. 4, a schematic structural diagram of an image capturing apparatus provided in an embodiment of the present disclosure is shown, where the image capturing apparatus may include the following modules:
a merchant main map obtaining module 310, configured to obtain a plurality of dish images of a merchant and a merchant main map;
a title text obtaining module 320, configured to obtain a title text of a dish corresponding to each dish image and a main picture title text corresponding to the merchant main picture;
a vegetable main map feature obtaining module 330, configured to obtain a vegetable feature in each vegetable image and a main map feature in the merchant main map;
a dish characterization vector determination module 340, configured to determine a dish characterization vector for each dish image based on the dish title text and the dish features of each dish image;
a principal graph characterization vector determination module 350 for determining a principal graph characterization vector for the merchant principal graph based on the principal graph title text and the principal graph features of the merchant principal graph;
the image cluster generating module 360 is configured to perform clustering processing on the plurality of dish images based on each dish characterization vector to generate a plurality of image clusters;
the estimated click rate obtaining module 370 is configured to obtain an estimated click rate corresponding to each dish image based on each dish characterization vector, the main graph characterization vector and a preset click rate estimation model;
and the to-be-displayed image acquisition module 380 is configured to acquire, from each image cluster, a dish image with the highest estimated click rate as the to-be-displayed dish image.
The image obtaining device provided by the embodiment of the disclosure obtains a plurality of vegetable images and a main merchant main map of a merchant, obtains a vegetable title text corresponding to each vegetable image and a main map title text corresponding to the merchant main map, obtains vegetable features in each vegetable image and main map features in the merchant main map, determines a vegetable characterization vector of each vegetable image based on the vegetable title text and the vegetable features of each vegetable image, determines a main map characterization vector of the merchant main map based on the main map title text and the main map features of the merchant, performs clustering processing on the plurality of vegetable images based on the vegetable characterization vectors to generate a plurality of image clusters, obtains a click rate corresponding to each vegetable image based on each vegetable characterization vector, the main map characterization vector and a preset click rate prediction model, obtains a vegetable image with the highest predicted click rate from each image cluster, as an image of the dish to be displayed. According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image type clusters, and the dish images with the highest estimated click rate are selected from each image type cluster to serve as the displayed dishes of the merchants, so that the redundancy of the displayed dishes can be avoided, and the use experience of users can be improved.
Example four
Referring to fig. 5, a schematic structural diagram of an image capturing apparatus provided in an embodiment of the present disclosure is shown, where the image capturing apparatus may include the following modules:
a merchant main map obtaining module 410, configured to obtain a plurality of vegetable images of a merchant and a merchant main map;
a title text obtaining module 420, configured to obtain a title text of a dish corresponding to each dish image and a main picture title text corresponding to the merchant main picture;
a vegetable main map feature obtaining module 430, configured to obtain a vegetable feature in each vegetable image and a main map feature in the merchant main map;
a dish characterization vector determination module 440, configured to determine a dish characterization vector for each of the dish images based on a dish title text and a dish feature of each of the dish images;
a principal graph characterization vector determination module 450 for determining a principal graph characterization vector for the merchant principal graph based on the principal graph title text and the principal graph features of the merchant principal graph; (ii) a
An image cluster generating module 460, configured to perform clustering processing on the plurality of dish images based on each dish characterization vector to generate a plurality of image clusters;
the estimated click rate obtaining module 470 is configured to obtain an estimated click rate corresponding to each dish image based on each dish characterization vector, the main graph characterization vector and a preset click rate estimation model;
the to-be-displayed image acquisition module 480 is configured to acquire, from each image cluster, a dish image with the highest estimated click rate as the to-be-displayed dish image;
and the display image generation module 490 is configured to perform fusion processing on the merchant main graph and each to-be-displayed dish image to generate a display image of the merchant on a specified service platform.
Optionally, the dish characterization vector determination module 440 includes:
the first characterization vector acquisition submodule 441 is configured to acquire a first characterization vector corresponding to a dish feature in each dish image;
a first text vector obtaining sub-module 442, configured to obtain, based on each of the dish title texts, a first text characterization vector corresponding to each of the dish images;
and the dish characterization vector generation sub-module 443 is configured to combine the first characterization vector of each dish image with the first text characterization vector to generate a dish characterization vector of each dish image.
Optionally, the master characterization vector determination module 450 includes:
a second token vector obtaining sub-module 451, configured to obtain a second token vector corresponding to the main feature;
the second text vector obtaining sub-module 452 is configured to obtain, based on the main picture title text, a second text characterization vector corresponding to the merchant main picture;
a principal graph token vector generation sub-module 453 is configured to combine the second token vector and the second text token vector to generate the principal graph token vector.
Optionally, the image class cluster generating module 460 includes:
an image cluster generation submodule 461, configured to perform clustering processing on the plurality of dish characterization vectors by using a preset clustering algorithm, so as to generate a plurality of image clusters; each image cluster comprises at least one dish image.
Optionally, the estimated click rate obtaining module 470 includes:
a similarity obtaining submodule 471, configured to obtain a similarity corresponding to each dish image based on each dish characterization vector and the main picture characterization vector;
the estimated click rate output sub-module 472 is configured to input each of the similarities to the preset click rate evaluation model, and output an estimated click rate corresponding to each of the dish images.
Optionally, the similarity obtaining submodule 471 includes:
and the cosine similarity operator module is used for calculating cosine similarity between each dish characterization vector and the main picture characterization vector and taking each cosine similarity as the similarity of each dish image.
The image obtaining device provided by the embodiment of the disclosure obtains a plurality of vegetable images and a main merchant main map of a merchant, obtains a vegetable title text corresponding to each vegetable image and a main map title text corresponding to the merchant main map, obtains vegetable features in each vegetable image and main map features in the merchant main map, determines a vegetable characterization vector of each vegetable image based on the vegetable title text and the vegetable features of each vegetable image, determines a main map characterization vector of the merchant main map based on the main map title text and the main map features of the merchant, performs clustering processing on the plurality of vegetable images based on the vegetable characterization vectors to generate a plurality of image clusters, obtains a click rate corresponding to each vegetable image based on each vegetable characterization vector, the main map characterization vector and a preset click rate prediction model, obtains a vegetable image with the highest predicted click rate from each image cluster, as an image of the dish to be displayed. According to the method and the device for displaying the dish images, the dish images of the merchants are divided into the plurality of image type clusters, and the dish images with the highest estimated click rate are selected from each image type cluster to serve as the displayed dishes of the merchants, so that the redundancy of the displayed dishes can be avoided, and the use experience of users can be improved.
An embodiment of the present disclosure also provides an electronic device, including: a processor, a memory and a computer program stored on the memory and executable on the processor, the processor implementing the image acquisition method of the foregoing embodiments when executing the program.
Embodiments of the present disclosure also provide a computer-readable storage medium, in which instructions, when executed by a processor of an electronic device, enable the electronic device to perform the image acquisition method of the foregoing embodiments.
For the device embodiment, since it is basically similar to the method embodiment, the description is simple, and for the relevant points, refer to the partial description of the method embodiment.
The algorithms and displays presented herein are not inherently related to any particular computer, virtual machine, or other apparatus. Various general purpose systems may also be used with the teachings herein. The required structure for constructing such a system will be apparent from the description above. In addition, embodiments of the present disclosure are not directed to any particular programming language. It is appreciated that a variety of programming languages may be used to implement the teachings of the embodiments of the present disclosure as described herein, and any descriptions of specific languages are provided above to disclose the best modes of the embodiments of the present disclosure.
In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the disclosure may be practiced without these specific details. In some instances, well-known methods, structures and techniques have not been shown in detail in order not to obscure an understanding of this description.
Similarly, it should be appreciated that in the foregoing description of exemplary embodiments of the disclosure, various features of the embodiments of the disclosure are sometimes grouped together in a single embodiment, figure, or description thereof for the purpose of streamlining the disclosure and aiding in the understanding of one or more of the various inventive aspects. However, the disclosed method should not be interpreted as reflecting an intention that: that is, claimed embodiments of the disclosure require more features than are expressly recited in each claim. Rather, as the following claims reflect, inventive aspects lie in less than all features of a single foregoing disclosed embodiment. Thus, the claims following the detailed description are hereby expressly incorporated into this detailed description, with each claim standing on its own as a separate embodiment of an embodiment of this disclosure.
Those skilled in the art will appreciate that the modules in the device in an embodiment may be adaptively changed and disposed in one or more devices different from the embodiment. The modules or units or components of the embodiments may be combined into one module or unit or component, and furthermore they may be divided into a plurality of sub-modules or sub-units or sub-components. All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and all of the processes or elements of any method or apparatus so disclosed, may be combined in any combination, except combinations where at least some of such features and/or processes or elements are mutually exclusive. Each feature disclosed in this specification (including any accompanying claims, abstract and drawings) may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise.
The various component embodiments of the disclosure may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It will be understood by those skilled in the art that a microprocessor or Digital Signal Processor (DSP) may be used in practice to implement some or all of the functions of some or all of the components in a motion picture generating device according to an embodiment of the present disclosure. Embodiments of the present disclosure may also be implemented as an apparatus or device program for performing a portion or all of the methods described herein. Such programs implementing embodiments of the present disclosure may be stored on a computer readable medium or may be in the form of one or more signals. Such a signal may be downloaded from an internet website or provided on a carrier signal or in any other form.
It should be noted that the above-mentioned embodiments illustrate rather than limit embodiments of the disclosure, and that those skilled in the art will be able to design alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses shall not be construed as limiting the claim. The word "comprising" does not exclude the presence of elements or steps not listed in a claim. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of the disclosure may be implemented by means of hardware comprising several distinct elements, and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by one and the same item of hardware. The usage of the words first, second and third, etcetera do not indicate any ordering. These words may be interpreted as names.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
The above description is only for the purpose of illustrating the preferred embodiments of the present disclosure and is not to be construed as limiting the embodiments of the present disclosure, and any modifications, equivalents, improvements and the like that are made within the spirit and principle of the embodiments of the present disclosure are intended to be included within the scope of the embodiments of the present disclosure.
The above description is only a specific implementation of the embodiments of the present disclosure, but the scope of the embodiments of the present disclosure is not limited thereto, and any person skilled in the art can easily conceive of changes or substitutions within the technical scope of the embodiments of the present disclosure, and all the changes or substitutions should be covered by the scope of the embodiments of the present disclosure. Therefore, the protection scope of the embodiments of the present disclosure shall be subject to the protection scope of the claims.